anguyen8/deepface-emd

Code for CVPR 2022 https://arxiv.org/abs/2112.04016 DeepFace-EMD Re-ranking Using Patch-wise Earth Movers Distance Improves Out-Of-Distribution Face Identification

27
/ 100
Experimental

This project helps improve the accuracy of face identification systems, especially when faces are partially obscured (like with masks or sunglasses) or cropped in unusual ways. It takes existing face images and a query image, then intelligently re-ranks potential matches to provide more reliable results. It's designed for computer vision researchers, forensic analysts, or security professionals who need to identify individuals from challenging images.

No commits in the last 6 months.

Use this if you need to boost the accuracy of face identification from 'out-of-distribution' images (e.g., partially masked, wearing sunglasses, or oddly cropped faces) without retraining your entire face recognition model.

Not ideal if you are looking for a standalone, production-ready face identification system rather than a method to enhance an existing one's robustness.

face-identification biometrics forensic-imaging security-systems image-retrieval
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 8 / 25
Maturity 8 / 25
Community 11 / 25

How are scores calculated?

Stars

55

Forks

6

Language

Python

License

Last pushed

Mar 29, 2023

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/anguyen8/deepface-emd"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.